The present invention relates to plasma processing systems for use in the manufacture of semiconductor integrated circuits. More particularly, the present invention relates to improved plasma processing system IPMs (integrated power modules) that offer improved reliability and lower acquisition/maintenance costs.
Plasma processing systems have long been employed in the manufacture of semiconductor devices (such as integrated circuits or flat panel displays). In a typical plasma processing system, the substrate (e.g., the wafer or the glass panel) is typically disposed inside a plasma processing chamber for processing. Energy in the form of AC, DC, RF, or microwave is then delivered to the plasma processing chamber to form a plasma out of supplied etchant or deposition source gases. The plasma may then be employed to deposit a layer of material onto the surface of the substrate or to etch the substrate surface.
As the electrodes require energy to ignite and sustain the plasma, a power delivery system is typically required to condition the AC power obtained from the grid, to transform the AC power into the appropriate form of energy required to ignite and sustain the plasma, and to provide the DC voltages for operating the control electronics.
To facilitate discussion, FIG. 1 illustrates a simplified power delivery system of a currently available plasma processing system known as the 4520XL.TM., available from Lam Research Corporation of Fremont, Calif. In the example of FIG. 1, plasma processing system 100 represents a parallel plate, multiple frequencies plasma processing system. It should be appreciated, however, that the discussion herein is not limited to this specific type of plasma processing system. In fact, the concept discussed herein is applicable to plasma processing systems in general irrespective of the number of electrodes, the geometry of the chamber, or the type of energy source employed. Further, although only one chamber of plasma processing system 100 is shown to facilitate discussion, it should be appreciated that a plasma processing system may take the form of a cluster tool, which may include one or multiple modules, each of which may have one or multiple chambers per module.
Referring now to FIG. 1, wafer 102 is shown disposed in a plasma processing chamber 104 for processing. More specifically, wafer 102 is shown disposed on a chuck 106, which acts as one electrode. The other electrode 108 is disposed above wafer 102 as shown. RF generator 110 represents a 27 MHz RF generator, which supplies RF energy to match a network 114 through a coax cable 122. As is well known, one function of the match network is to match the impedance of the plasma to that of the generator in order to maximize power delivery. From match network 114, the RF energy is provided to electrode 108 through a diplexer 118. A diplexer is a well known device that passes energy of a certain frequency while passing energy having other frequencies to ground. Since electrode 108 is a 27 MHz electrode, diplexer 118 passes 27 MHz RF energy to electrode 108 while passing RF energy having other frequencies to ground.
Likewise, RF generator 112 represents a 2 MHz RF generator which supplies the RF energy to match network 116 through coax cable 124. From match network 116, the RF energy is supplied to a diplexer 120 through coax cable 126. Diplexer 120 passes 2 MHz RF energy to chuck 106 and passes RF energy having other frequencies directly to ground.
Nowadays, the various major functional blocks of a power delivery system (e.g., generators, matches, diplexers, or the like) are typically distributed among multiple subsystems, many of which are enclosed in their own EMI enclosures and include their own DC power supplies. This is because the current practice in power delivery system design is to render the major functional blocks or subsystems as modular as possible. In other words, the current practice is to provide each subsystem with sufficient local resources onboard (e.g., DC power supplies to operate the local electronics) so as to enable a given subsystem to be readily adapted for use in a plug-and-play fashion in many different plasma processing systems. By commoditizing these subsystems, the vendors of these subsystems hope to achieve economy of scale since fewer subsystems need to be designed and inventoried for the plasma processing equipment market.
There is also another design philosophy in the semiconductor processing equipment industry which favors the provision of resources required by each subsystem (e.g., DC power supplies) in the subsystems themselves. As plasma processing systems become more complex and expensive, lower cost of ownership is achieved by reducing the amount of time that the plasma processing system is out of service due to equipment failures. Beside improving the quality of the subsystems, vendors of plasma power delivery systems believe that by distributing the resources among the various modular subsystems, the effects of a subsystem failure can be isolated and addressed quickly. By making the subsystems modular and self-sufficient in terms of required resources, the failed subsystem can be swapped out, and the plasma processing system can be brought back into operation quickly.
As a practical matter, each of these modular subsystems (e.g., match networks 114 and 116, diplexers 118 and 120 and RF generators 110 and 112) occupies a nontrivial amount of space. Accordingly, it is oftentimes impractical to position these subsystems close to the plasma processing chamber and still provide adequate space for maintenance. The crowding problem is exacerbated in a cluster tool environment where multiple chambers may be positioned in close proximity to one another.
In the prior art, the crowding problem is addressed by moving certain subsystems to a remote location and to connect the subsystems together via conductors/or and coax cables. With reference to FIG. 1, for example, RF generators 110 and 112, along with their water cooling systems and control electronics, may be positioned away from the plasma processing chamber to relieve crowding. In the typical case, RF generators 110 and 112 may be installed on a rack some distance away (50-60 feet in some cases) from the plasma processing chamber. Other subsystems such as matches and/or diplexers may be located closer to chamber 104 within the assembly shown as plasma processing module 150. Coax cables 122 and 124 are then employed to couple the RF generators on rack 152 to the subsystems at plasma processing module 150.
Because the subsystems of the power delivery system are now split among multiple locations, separate power distribution boxes are required. With reference to FIG. 1, rack 152 requires a power distribution box 154 to receive AC power from the grid (e.g., in the form of 208 volts, 3-phase) and to distribute AC power to RF a generators 110 and 112 via conductors 156 and 158. These conductors 156 and 158 plug into RF generators 110 and 112, which are provided with complementary plugs for quick connection and disconnection. Generator 110 also includes an additional connector for connecting with coax cable 122 (which supplies the RF energy to match network 114). Likewise, RF generator 112 also includes an additional connector to couple with coax cable 124 (which supplies the RF energy to match network 116).
DC voltages to the control electronics within RF generators 110 and 112 are provided by DC generators, which are typically provided onboard each RF generator to satisfy modular design guidelines. In the example of FIG. 1, RF generator 110 is shown having a DC power supply 162 for converting the AC voltage received at RF generator 110 to the DC voltages levels required by its control electronics. Likewise, RF generator 112 is shown having a DC power supply 164 for converting the AC voltage received at RF generator 112 to the DC voltage levels required by its control electronics. If other sensors or control electronics external to the generators exist on rack 152, additional DC power supplies may be provided on rack 152. For safety, rack 152 typically comes with an EMO (Emergency Off) subsystem 160, which is essentially a panic switch that allows AC power to be shut off quickly in case of emergency.
Since plasma processing module 150 is remoted from rack 152, a separate power distribution box is now required for plasma processing module 150. This power distribution box is shown as power distribution box 170, which includes its own separate DC power supply 172 to provide the requisite DC voltages to the control electronics local to plasma processing module 150. Like power distribution box 154, power distribution box 170 includes the contactors, relays, and connectors required to provide AC power to the various AC loads of plasma processing module 150 such as the pumps, heaters, and the like. Conductors from DC power supply 172 lead to various control electronics (including those on the chamber and in the match networks), sensors, and other DC loads within plasma processing module 150. For safety reasons, an EMO (Emergency Off) subsystem 174 is provided with plasma processing module 150 to permit power to the subsystems of plasma processing module 150 to be shut off in an emergency.
It has been recognized by the inventor herein that the current power delivery system has certain disadvantages. By way of example, the attempt to turn each major functional block of the power delivery system (such as generators, matches, diplexers, or the like) into a modular, stand-alone subsystem introduces unnecessary redundancy of components into the assembled power delivery system. This is because there are multiple DC power supplies, EMO subsystems, power distribution boxes, power line filters, relays, connectors, contactors, coax cables, conductors in the power delivery system after it is assembled from the modular subsystems. The high component count of the assembled power delivery system disadvantageously increases the acquisition cost.
Ironically, the redundancy of components does not increase reliability since when a component fails, the subsystem affected would still bring down the entire plasma processing system since these redundant components exist in different subsystems and do not serve as backups for one another. In fact, the high component count decreases reliability since there are now more components to fail.
As mentioned earlier, the attempt to modularize the major functional blocks of the power delivery subsystem also renders it difficult to package these subsystems, each with its own bulky EMI enclosure, within the tight space available around each plasma processing chamber. When certain subsystems are positioned away from the plasma processing chamber to relieve crowding, long coax cables and conductors are required, which introduces losses into the power delivery system and increases EMI concerns. The use of coax cables and coax connectors for providing high voltage/high current signals between the various subsystems (e.g., between an RF generator and its match network) also decreases reliability. This is because the coax cables and/or the connectors tend to break down over time when they are required to carry high voltage/high current signals.
In view of the foregoing, there are desired improved power delivery systems for providing power to ignite or sustain plasma in a plasma processing chamber.